DangerousFat said:

That's cool, you work on anything interesting or any multi-player strategy games?

I think you're not understanding the model I presented or the methods I used to come up with it. First of all, the model isn't complete and doesn't claim to be. The model only predicts ship squad building points based on stats. It doesn't say anything about movement dials, action bars, upgrades, pilot special abilities, or anything else.

It seems like you don't understand what a regression analysis is or how to read one. It isn't a creation tool. It can't be used to make a new game system. Ultimately, the formula the game designers used to come up with ship squad values was arbitrary and derived from intuition and testing. Regression analysis can only show what the underlying formula is, not why it was chosen. I don't know why each attack or defense die above 2 costs 8 squad points instead of any other number. I just know it does. The odds that it doesn't are infinitesimally small (1 in many trillion). That is shown in the regression analysis table in the 't stat' and 'p-value' columns. If you don't understand stats (and by 'stats' I don't mean something as banal as pop culture stat factoids), then you can't understand the significance of those numbers and will probably dismiss what I'm saying.

Also, if you read through the whole post (which I don't think you did, no offense), you will see that I agree with your last point. The formula doesn't fit perfectly and I speculate the deviations (Academy Pilots and most named pilots, specifically) were added for game balance reasons.

Often times points are added or subtracted to influence game play. My experience so far is that maneuvering is still the most important factor.  Which, of course, I'm not sure how you could mathematically balance that in and of itself without serious playtesting.  I feel as if that may be what some of that mysterious initial cost may be attempting to influence. 

 Wow.  Wierd to do all that for a game that involves pretending to shoot your plastic toy ships with my plastic toy ships.

 I would rather just play the game with some beer and friends than spend any time whatsoever dissecting it to death.

….or spend that time unnecessarily bashing someone for being interested in doing something that you're not, apparently.  I'd say that's less healthy than making a spreadsheet. 

Xnath said:

You could say that about any game. That's hardly an enlightened, unique, or even interesting point of view. Classy use of a beer reference too; that really added to your comment.

Xnath said:

I'm sorry i'm asian, it's like an instinctual, innate kinda inclination you know? That, and high expectations asian father will kung fu you to death if you did not get A+ for your exams and a 100-win streak in star wars miniatures

as always interesting and as people have said no formula is perfect but interesting exercise all the same although i have to admit this level of maths scares me

Hardrainfalling said:

Excel did all the hard/tedious math. It really wasn't that bad. It is all in knowing how to set it up and then knowing how to read it.

ShadowJak said:

This is why I'm jealous of you Shadow….I don't remember my college math and am not very comfortable with Excel…so I just used the brute force method by trying tons of different numbers on scratch paper.  After like the 50th failure I was cursing myself for not having a more intelligent way to do it lol.

You mentioned constants.  Way I see it, in addition to the above, each ship also starts off with universal/faction specific abilities (Focus for all and either Target Lock for Rebels or Evade/Barrel Roll for Imperials).

 Then, after a certain number of points, the Pilot selects a tactic/special ability that they can do.  This represents them learning how to use their fighter better instead of just an increase to Pilot Skill.

What do you think?

Sarone said:

No. Not at all.

Well, I happen to know something about Linear Regression and its not the be-it-all mathematical tool. Yes, game designers use some sort of mathematical model to cost stuff, but quite often in games the "extremes" tend to be a lot better. Would a 0 defense and 1 hull ship with 5 attack and 4 shields be very good or very bad?

Linear Regression can give some awnsers, but linear regression as some serious problems if the extremes tend to deviate (either because they are a LOT better or a LOT worse). And ofcourse the second problem is the lack of data points and combinations (have we seen a ship with only 1 attack?). However, the "deviation from baseline" approach does tend to work; question would be if the baseline for the Imperial is the same one as for the Rebels. I have a hard time believing that, as point costs are usually relative to the type of army you can field. Rebels are slower. less defense but have shields and attack. Imperials are faster and have more defense, but have less attack and less shields.

I've played plenty of Warhammer 40k where costing units is quite hard. If you then compare one army with another army you tend to see discrepancies because the baseline of the armies are different baselines. Large deviation from that baseline tends to cost a LOT more than small deviation (attack 4 would cost less for the rebels than attack 4 would cost the Imperials, assuming the baseline is the ships provided in the base set). Moving a bit more when you are normally quite slow would cost more than moving a bit more when you are normally quite fast. Etc. In D&D for example getting a +4 bonus on a Stat would raise the cost by almost 50% compared to getting only a +3 bonus.

In these cases, Linear Regression would not yield anything UNLESS you modify the inputs to account for this, for example by subtracting the actual value from the baseline and then squaring the number if positive or taking it as the exponent to a power. Ofcourse, there are loads of options to play with using these mathematical tools but for games this might be overkill. For games, playtesting lots of games and testing out the values as well as the extremes will be at least more fun than the mathematical way, but you do need to know your statistical probabilities (what is the chance of a critical hit inflicting two damage? What is the chance of scoring at least 3 hits with 3 dice?).

Very interesting thread indeed!

I decided to play around a bit and see what numbers I could come up with.  First I vetted my excel to match the original post's and my values seemed to match up.

I took Steven's specs for the CR-90 Corvette to see how many points it'd cost using your formula.  Here's what I came up with:

Pilot = 0

Weapons = 8

Agility = 1

Hull = 10

Shields = 10

Cost = 116.75!!!

I also decided to see what the Z-95-AF4 Headhunter would estimate to since it seems that people want it to be added as an expansion.  Here's what I came up with:

Pilot = 4

Weapons = 2

Agility = 2

Hull = 2

Shields = 1

Cost = 6.25 (Rebel swarm tactics anyone?)

I justfied my values from this quote from Wookiepedia: "When compared to the T-65, it was slower, less maneuverable, had lighter armor and shielding, and was not as heavily armed.

marcelvdpol said:

What you have pointed out is that linear regressions don't work well for functions that are not linear.

marcelvdpol said:

"Well, I happen to know something about Linear Regression and its not the be-it-all mathematical tool. Yes, game designers use some sort of mathematical model to cost stuff, but quite often in games the "extremes" tend to be a lot better. Would a 0 defense and 1 hull ship with 5 attack and 4 shields be very good or very bad?"

I never claimed it was. It appears you didn't read my post. The point of the model isn't to determine if certain stats are "good or bad." The point of the model is to show how the ships' squad building points are determined.

"Linear Regression can give some awnsers, but linear regression as some serious problems if the extremes tend to deviate (either because they are a LOT better or a LOT worse). And ofcourse the second problem is the lack of data points and combinations (have we seen a ship with only 1 attack?). However, the "deviation from baseline" approach does tend to work; question would be if the baseline for the Imperial is the same one as for the Rebels."

You would have seen that the baseline appears to be same, if you had read my post.

"I have a hard time believing that, as point costs are usually relative to the type of army you can field. Rebels are slower. less defense but have shields and attack. Imperials are faster and have more defense, but have less attack and less shields."

Incredibly facile. It seems like you didn't read all of my post or if you did, then you didn't understand it.

"I've played plenty of Warhammer 40k where costing units is quite hard. If you then compare one army with another army you tend to see discrepancies because the baseline of the armies are different baselines. Large deviation from that baseline tends to cost a LOT more than small deviation (attack 4 would cost less for the rebels than attack 4 would cost the Imperials, assuming the baseline is the ships provided in the base set). Moving a bit more when you are normally quite slow would cost more than moving a bit more when you are normally quite fast. Etc. In D&D for example getting a +4 bonus on a Stat would raise the cost by almost 50% compared to getting only a +3 bonus."

X-Wing has simple stats compared to most other miniature games. Also, there are only 4 ship types so far. If you had read my post you would have seen there is a base line. Just because other games have different baselines between factions doesn't mean that X-Wing does.

"In these cases, Linear Regression would not yield anything UNLESS you modify the inputs to account for this, for example by subtracting the actual value from the baseline and then squaring the number if positive or taking it as the exponent to a power. Ofcourse, there are loads of options to play with using these mathematical tools but for games this might be overkill. For games, playtesting lots of games and testing out the values as well as the extremes will be at least more fun than the mathematical way, but you do need to know your statistical probabilities (what is the chance of a critical hit inflicting two damage? What is the chance of scoring at least 3 hits with 3 dice?)."

If you had read my post, you would have seen that I did modify the inputs.

If you had read my post, you would have seen where I mentioned the fact that attack dice are better than defense dice.

The probabilities of attack vs. defense are outside the scope of this post and don't matter because they aren't accounted for in squad building points. if you had any understanding of what I posted, you'd already know that.

I know the attack and defense probabilities. I know the probabilities when taking into account Focus, Target Lock, and/or Evade. I even know how well missiles and torpedoes stack up against normal attacks. I could make a post about it but won't because it isn't interesting; most of it is entirely intuitive (more dice is more damage, hurr durr).

I'm not trying to be overly harsh but taking a stats 101 class isn't experience with linear regression. Knowing how to click the buttons to do a regression in excel isn't the same as understanding it. Nothing you said was insightful in the least. Everything you brought up has already been addressed. You obviously don't even know what the t stat or p-value mean. If you did, you wouldn't have said half the things you said. If you had read the post and the rest of the comments in the thread, you wouldn't have said the other half of the things you said.